Supply chain is a good example of this. For decades, more information has been available than companies could actually use. POS (point-of-sale) data has been available for 40 years. The ability to scan a product and immediately log the sale was a breakthrough, but actually using that information to impact business practices for manufacturers was impossible for decades. A large consumer products company will ship more than 1000 individual SKUs (stock keeping units) to 5000 retail sites in the US alone. That is a lot of information to analyze and it wasn’t possible – back then.
Advances in computer power have reached a point where using this data to improve the supply chain is possible, but not if manufacturers continue to rely on the same old mathematical techniques they have always used. Now that technology has reached an advanced state, companies must look for new mathematical approaches. For example, manufacturers have relied on time-series statistics to forecast sales. Times-series models analyze past observations of purchases and shipments to forecast future purchasing behavior. The problem is the results are just seasonal patterns and trends, ignoring important events like promotions and competitive activity. If a manufacturer publishes a coupon in the Sunday paper then sales for the following week will be higher because of the marketing activity. With historical data alone, the variables above are not accounted for.
Developing a forecasting engine specific to supply chain business realities overcomes the inherent limitations of time-series models. There is no need for manual overrides to the forecast and reworking the math so exceptions fit the model. The new model is designed to analyze all available demand data, including POS, retailer inventory, inventory currently moving through the supply chain, orders, shipments in route, and the historical forecast. All of these pieces of data are inputs into the new mathematical model, all used in the equation to determine what future demand will be. The historical forecast is just one piece of the puzzle, and all other influencers are analyzed, evaluated and used to generate a new forecast of demand that is frequently 40 percent more accurate.
Such math models are not unique to the supply chain industry or consumer products manufacturers – Amazon, Netflix Recommendations and iTunes Genius are examples of companies successfully developing mathematical models to provide superior customer service. When logging onto Amazon.com, shoppers are greeted with recommendations created for the individual based on past purchases. After a purchase is made, shoppers are shown a list of similar purchases others made with a suggestion that individuals who purchased book A also purchased book B. The website will even ask for additional input on books already owned or purchased, whether the feedback is positive or negative and create an even better list. These recommendations are based on mathematical models designed specifically for Amazon.com to accumulate and evaluate data for its customers. Similarly, when ordering a movie from Netflix, the company provides a list of recommended movies based on previous experiences. Perhaps the best example is Apple’s iTunes Genius, a program designed specifically to create a list of recommended songs by taking previously selected songs and creating a customer-specific list of 25 songs.
Computers have the ability to improve all aspects of our lives, but only if we take advantage of what they offer. Relying on outdated mathematical models to solve business problems when new computers have the capacity to do so much more is a losing proposition. Business innovators and industry leaders know the way to win in the marketplace is to create new and better ways of meeting customer needs.
Robert F. Byrne is president and CEO of Terra Technology, a provider of innovative supply chain solutions for consumer products companies. For more information, please visit www.terratechnology.com.